Speech to List
Overview
Speech to List lets Admin and Root users create a full task list by speaking. Tap the microphone button on the task list page, describe the work that needs doing in natural language, and Ziptask transcribes the recording and uses AI to convert it into a structured list — title, description, start and due dates, a repeat schedule, and individual task items (including which ones need a photo as proof) — ready to review, edit, and assign in one step.
Designed for field supervisors who are walking a job site, inspecting a property, or calling out work that needs doing — without needing to stop and type.
Requires:
speech_to_listfeature flag. Business plan only.
Permissions
| Action | User | Admin | Root |
|---|---|---|---|
| Access Speech to List (microphone button) | ✗ | ✓ | ✓ |
| Review and edit the generated list | ✗ | ✓ | ✓ |
| Save the generated list | ✗ | ✓ | ✓ |
User-role members do not see the microphone button.
How it works
- Tap the microphone button — the button pulses to show recording is active.
- Speak your tasks — describe the work in plain language. There is no need to use a special format; natural speech works best (see examples below). Recording stops automatically after 60 seconds or when you tap the button again.
- Processing — the recording is transcribed and sent to the AI, which extracts a list title, an optional description, start and due dates, a repeat schedule, and individual task items. This usually takes a few seconds.
- Review — a pre-filled review form opens with everything the AI picked up. You can edit any field, add or remove items, change the start/due dates, adjust the repeat schedule, mark which items need a photo, and set assignments, team, tags, project, and customer before saving.
- Save — the list and all its items are created in one step, exactly like any manually created list.
What Speech to List captures
The AI listens for the details you naturally say out loud and fills them in for you. Everything can be corrected in the review step before saving.
| Detail | What to say | Where it lands |
|---|---|---|
| Title | ”Get the east wing ready…” | List title |
| Description | Any framing before the tasks | List description |
| Start date | ”starting Monday”, “from next week” | Start date |
| Due date | ”by Friday”, “before the 15th” | Due date |
| Repeat schedule | ”every day”, “every weekday”, “every other week”, “every Tuesday and Thursday” | Recurrence (frequency + interval + days) |
| Task items | Each discrete action | Individual items |
| Photo proof | ”take a photo when it’s done”, “picture of each one” | Item marked “photo required” |
Some fields still need to be picked in the review step because they reference existing account records the AI can’t match by ear: assignee, team, tags, project, and customer. (Speech will try to match a spoken name to a member, but confirm it in the review step.) Choosing a project that already has a customer fills in and locks the customer to match — the same as the manual create form.
Sample recordings
The AI handles natural, conversational speech — you do not need to list items formally. Here are examples of the kind of recordings that work well:
Site walkthrough before a crew arrives:
“We need to get the east wing ready before the crew shows up tomorrow. First, clear the debris from the stairwell on floors 2 and 3. The emergency exit signs on floor 2 need new batteries — check all of them. Patch the drywall damage in the break room and make sure the fire extinguishers are tagged and in their mounts. Last thing — the loading dock light has been flickering, get someone to look at that.”
What gets generated: A list titled something like “East Wing Prep” with 5 items: stairwell debris, emergency exit sign batteries, drywall patch, fire extinguisher check, loading dock light.
Property turnover between tenants:
“Unit 4B needs to be ready by Friday for the new tenant. Patch the two holes in the bedroom wall and repaint — match the existing white. The bathroom faucet is dripping, replace it. Deep clean the kitchen including the oven and hood vent. Test all the outlets with a tester and replace the outlet cover in the hallway, it’s cracked. Check that the sliding door locks properly.”
What gets generated: A list titled something like “Unit 4B Turnover” with 6 items broken out per task.
End-of-day closing checklist:
“End of day checks before everyone leaves. Lock all the supply rooms on the ground floor. Turn off the compressors in bay 2 and 3 — they’ve been left on twice this week. Make sure the loading dock doors are sealed, both of them. Check the job log is updated for tomorrow’s crew and leave it on the foreman’s desk.”
What gets generated: A list titled something like “End of Day Closing” with 4 items matching the spoken tasks.
Weekly safety walkthrough:
“Monday morning safety check. Inspect the scaffolding on levels 2 and 3 — look for loose boards and missing clips. Verify all fall harnesses are back in the equipment room and not left on site. Check the first aid kit in the site office is fully stocked and the eyewash station is functional. Make sure the forklift horn and reverse alarm are working before it goes back on the floor.”
What gets generated: A list titled something like “Monday Safety Check” with 4 items.
Recurring rounds with dates and photo proof:
“Set up the daily bathroom rounds, starting Monday and repeating every weekday. Restock the paper towels and soap, and take a photo of each stall when it’s cleaned. Empty the bins and wipe down the sinks.”
What gets generated: A list titled something like “Daily Bathroom Rounds”, a start date of Monday, a repeat schedule of every weekday, and items broken out per task — with the stall-cleaning item marked “photo required.”
Tips for better results
- Speak in complete thoughts — “Replace the bathroom faucet” works better than just “faucet”.
- One topic per sentence — the AI treats each discrete action as a separate task item.
- Mention the location or context early — “Get unit 4B ready for Friday” gives the AI a clear title anchor.
- Say dates and cadence naturally — “starting Monday”, “due Friday”, “every other week”, or “every Tuesday and Thursday” all map to the schedule fields.
- Call out photo proof per task — “take a photo when it’s done” marks that item as requiring a photo before it can be submitted.
- Background noise is fine — the transcription handles typical outdoor/site noise, but try to avoid talking over machinery if possible.
- Review before saving — the generated items are a starting point. Add, remove, or rename items freely before saving.
Support notes
- Microphone button not visible: Confirm
speech_to_listis enabled in Account → Settings → Feature Flags and that the user’s role is Admin or Root. The button is hidden for User-role members. - “Microphone access denied” error: The device or browser has blocked microphone access. On mobile, go to device Settings → Ziptask (or the browser app) → Permissions → Microphone and enable it. On desktop, check the browser’s site permissions for the Ziptask URL.
- Processing failed / no list generated: This typically means the recording was too short, too noisy, or contained no recognizable task descriptions. Try again with a clearer recording of at least 5–10 seconds.
- Items look wrong or are missing: The AI is not perfect — it works best with clear, action-oriented language. Use the review form to correct any mistakes before saving. Nothing is saved until the member taps Save.
- Recording cut off early: Recordings are capped at 60 seconds. If more tasks are needed, save the first list and start a second recording.